Hi,

My AD8310 log amp circuit only differs from the basic connection shown in the data sheet by having 200pF connected to OFLT pin, 680nF input coupling, and I have connected the dc Vout pin4 directly to Vin+ of a 22bit ADC ( not ADI ).

My circuit is built on double sided pcb, ground plane both sides, smt components and recommended decoupling; however no metal box is used for shielding/screening.

I use input matching for measuring power in a 50Ohm circuit with the recommended parallel 52R3 across the input; good for LF, HF frequencies. But for higher frequencies, VHF and up, I am not using any high pass input matching network .

I have calibrated my circuit using a 2-point calibration at 4MHz using 0dBm and -20dBm inputs; slope 24.57mV/dB, intercept -95.37dBm

When I made measurements with inputs set to -10dBm and -20dBm in the frequency range 10MHz-450MHz the output signal reduced by up to 10dB; please see attached graph.

This was completely unexpected and I was very surprised, as the data sheet contains the following statements viz, '-3dB bandwidth 900MHz', 'logarithmic linearity typically within +/- 0.4dB up to 100MHz' and '+/- 0.4dB linearity DC-400MHz'

So what is going on ? Why is the response not flatter ?

Has anyone got round this problem and have any suggestions ?#

I suspect you are getting such erroneous results because you are using the 4MHz calibration (slope and intercept at 4 MHz) to calculate the input power all the way up to 450 MHz. Slope and Intercept change quite a lot over frequency (please see figures 4, 5, 16 and 19 in the datasheet for an examples of slope and intercept shift over frequency).

You will need to calibrate at each frequency of interest to calculate the correct input power.

The linearity numbers you quote above are with referenced to 25C data at each frequency. So for example, you could get +/- 0.4 dB @ 50 MHz, if you applied input power sweep at 50 MHz and measured all the corresponding VOUT voltages. Take this VOUT vs. Input Power data and generate slope and intercept numbers. Finally, the error can be calculated as the difference between the actual VOUT data, and an ideal line generated with the 50 MHz slope and intercept you've calculated.

The -3 dB 900 MHz BW number quoted above is for each individual (non-linear) amplifier; there are six cascaded limiting amps that constitutes the limiting amplifier chain. These amplifiers interact with one another and the 1) overall bandwidth is not 900 MHz anymore, 2) the BW isn't necessarily flat, and 3) the nine detector cells can affect the flatness vs. frequency as well (remember, the limiting amps are only part of what makes the overall function of the logarithmic amplifier).